Ada-Instruct: Adapting Instruction Generators for Complex Reasoning
This addresses a bottleneck in instruction augmentation for complex reasoning tasks like code completion, offering an incremental improvement over prior methods.
The paper tackled the problem of generating complex instructions for large language models, showing that fine-tuning open-source LLMs with only ten examples can produce instructions of length ≥100, which existing Self-Instruct methods fail to do.
Instructions augmentation is a crucial step for unleashing the full potential of large language models (LLMs) in downstream tasks. Existing Self-Instruct methods primarily simulate new instructions from a few initial instructions with in-context learning. However, our study identifies a critical flaw in this approach: even with GPT4o, Self-Instruct cannot generate complex instructions of length $\ge 100$, which is necessary in complex tasks such as code completion. To address this issue, our key insight is that fine-tuning open source LLMs with only ten examples can produce complex instructions that maintain distributional consistency for complex reasoning tasks. We introduce Ada-Instruct, an adaptive instruction generator developed through fine-tuning. We empirically validated Ada-Instruct's efficacy across different applications. The results highlight Ada-Instruct's capacity to generate long, intricate, and distributionally consistent instructions.